1. Field of the Invention
The present invention relates to an accident sound detector for efficiently detecting traffic accidents from accident sounds (including accident sounds such as crash sounds and the sound of hard braking leading to an accident) at an intersection.
2. Description of the Prior Art
Generally, traffic accidents frequently occur at intersections, a traffic accident situation recorder that records the conditions when a traffic accident occurs has already been contemplated.
FIG. 3 shows a block diagram of a conventional traffic accident situation recorder as disclosed in Japanese Patent Application Laid-Open No. 4-338900.
Referring to FIG. 3, traffic signals 1, 4; signal lamps 1B, 1Y, 1R, 4B, 4Y, 4R; signal lines 2B, 2Y, 2R; a signal flasher 3; a signal lamp symbol generator 5; a time-of-day character generator 6; an adder 7; an imaging device 8; an endless recorder 9; an ambient sound detector 10; a crash sound detector 11; a record controller 12; and magnetic record reproduction device 13 are shown.
Here, a section for detecting traffic sound is constituted by the ambient sound detector 10 and the crash sound detector 11. That is, the ambient sound detector 10 detects a sound signal, and sends it to the crash sound detector 11. The crash sound detector 11 determines whether or not the detected sound signal is that of a crash between vehicles, and sends a crash sound detection signal to the record controller 12 when it determines that the detected sound signal is a crash sound.
The crash sound detector 11 makes this determination by comparing the level of the input sound signal with a predetermined threshold, and determines that there has been an accident when the level exceeds the threshold. In addition, to prevent erroneous operation, a directional microphone is used for the ambient sound detector 10, and a band-pass filter is installed in the path of the sound signal.
The traffic sound detection section described in the above application determines a crash sound by comparing levels of a detected sound signal with a predetermined, fixed threshold. However, ordinary traffic sounds includes many loud sounds other than accident sounds such as hard braking sounds, horn sounds, the sound of road repair, the sound of opening or closing shutters of surrounding stores, and the sound of speeding motorcycles.
Thus, with the method that compares the levels of a detected sound signal and a fixed threshold, various traffic sounds, other than actual accident sounds, are also determined to be accident sounds, thereby increasing the number of records that are not accidents.
In addition, such determination significantly depends on the size of the intersection, traffic, types of passing vehicles, number of surrounding stores, or evenness of the road, which leads to large differences in the number of records per unit hour from one intersection to another.
However, since the period of time available for recording on the recording medium is usually limited with a conventional device, the recording medium is exhausted in a short period of time if the number of records increases, and an image of an accident might be overwritten by successive records, so that necessary information can not be obtained. In addition, there is a problem in that consumption of the recording medium where the device is installed varies from one intersection to another. For example, in the case of a one-car accident, the driver might leave without reporting the accident. In such a case, it becomes impossible to identify the vehicle that caused the accident because information on the vehicle cannot be obtained.
In addition, in the case where sounds higher than the threshold are frequently generated, such as the sounds of road excavation or the noise of jack-hammers due to road repairs around an installed intersection, many records occur on the recorder in a short period of time. This is a large factor that increases the number of records.
Furthermore, since the level of sound becomes lower as the sound becomes farther from its source, for some intersections there is a problem in that the threshold level needs to be adjusted to an optimum level according to the distance between the sound source and the ambient sound detector.